320Kb/s MP3 vs. WAV

See the Dec 11th post on http://www.cdnav.com

I know, I know; all of you hardcore cats are scratching your heads and asking “why the hell would he even bother”, but please, bear with me guys, its something that I for one have always wished someone would hang on the web. If I was so damned curious, I suspect others may be as well…

The files are sourced from a high-resolution CD whose musical content is loaded with cymbals (which I gather are excellent in singling out smearing / ‘pumping’ in codecs), and created using the very latest software that I could get my hands on; LAME ver. 3.97 and EAC ver. 0.99 pb3, on a very nice audio-specific / tweaked, PC workstation. I also employed Audacity ver. 1.26 in some of the work.

Does anyone out there find these files reasonably easy to discern?

Andrew D.
www.cdnav.com

I THINK, I can speak for everyone when I say "NO".

You should do some extensive ABX testing, to find out if you are able to discern the encoded files from the source (as mentioned in the TOS).

If you are SO worried about quality and appearently don't mind the size, why not go for lossless? That will keep your mind clear.

Without listening very carefully to those files (and I don't intend to do so since I don't want to ruin my lossy comfort) I cannot distinguish between them.

For the record:
1. "High Resolution CDs" don't exist. CD audio is a fixed format, 44.100 Hz, 16 Bit.
2. "Audio-specific" workstations are not necessary at all. As long as the CD is read correctly, every properly working PC should write exactly the same audio data into the files.

In my experience I have had difficulty discerning a difference in many things if I listen to one then the other etc... However, after long periods of listening I find myself fatigued by 320kbps files. I ripped 1000 CDs to this format once, then re-ripped them all to AIFF and will never go back.

- Chris
Computer Audiophile | Turn Down The Silence

http://www.computeraudiophile.com

here, I made this web page today just for people like you.


http://www.m-ideas.com/sullivan/audio/woo.htm

Thats a nice joke there buddy :-)

Simple answer:

Lossy compression is never perfect. It just gets more and more "almost perfect". Most people dont notice the difference between "almost perfect" and "perfect" almost all the time. MP3 320kbit CBR is "almost perfect". If "almost perfect" isn't enough for you, then the only alternative is "lossless". No, there is no way to get both "absolute perfection" and "MP3-Filesize" at the same time.

Pre-echo effects are obvious.

(edit: did 3 more trials to get the p value lower)


Yes. Although certainly not for casual listening.

it will take some golder ears to ABX this... I mean, really golden...

benski

Educate me man - is it a pre-echo on cymbals, drum beats, or simply everything? I would like to know what to listen for. Also, what hardware are you using - headphones, speakers, PC or laptop, Mac etc? I am using an Asus Xonar sound card on a PC running Windows Vista via Audio-Technica ATH-M40fs headphones.

Always willing to learn.

Andrew D.

Fatigued by 320 kbps!!! That's absurd! Meaningless!

An easy artifact to focus on: about 12 seconds in, there is a snare hit. There is an attack at the beginning that disappears in the MP3 version. It's subtle, but if you relax and listen you'll notice it.

I was listening on a PC. Orban Optimod-PC 1100 soundcard (all processing turned off), balanced XLR output to an analog mackie mixer, Beyerdynamic DT 880 headphones.

Thank you.
Bookmarked, archived, and redundantly backed-up.

Golden usually means trained. You can easily "upgrade" your ears by training. Of course the physical limits of your hearing will remain (you likely won't hear 30 kHz e.g. no matter how hard you try).
Transparency (of a device or process) is not an absolute qualification. It can apply to a single person or to a group. It can vary over time (training, hearing loss e.g.), vary per person (different sensitivities), vary per source (problem samples) etc.
Has it ever been tried to specify the degree of transparency, e.g. in % of population or dB below threshold ? (like dB(A), V(RMS), THD+N(20Hz-20kHz))
Which attributes would be required to make the term transparency unambiguous ?

Kees de Visser

Always? With all tracks?

And even if the answer is 'Yes' for you, is it yes for everyone?

Extant evidence is that the vast majority of people who have tried to ABX high bitrate LAME mp3 vs wave, using various musical examples not already known to be 'killer', can't. That includes people who have reported their trials here...who could all be classified as 'casual' listeners.

So no, I would not claim that 'pre echo effects are 'obvious', even if they were to *me*.

the human hearing is very sensitive to pre-echo, in any case - when known or sought. Not all objects in audio pre-echo as audibly as other, or in certain contexts. It's all circumstancial.
Listening to music in almost any realistic environments, you shouldn't be able to hear these flaws unless they would be extremely obvious, ie. never.

Well, there is hybrid compression with WavPack. Granted, the file size of the lossy file and the correction file together become large, but there is still the high bitrate lossy part suitable for portable playback.

And, no, I do not train myself particularly hard to discern artifacts in lossly files so I can not hear any particularly well. Artifacts usually get lost in the noise the crowd makes anyways.

Sure there are problematic samples, people with golden ears and sometimes a few kinks with an encoder, but still I think it's irrelevant if other people are able to discern artefacts. If he thinks this may be a problem, he should perform some ABX tests to train his ears and see if he is able to discern these files, but still why not eliminate any possible problems entirely and use lossless?